JP2006271474A - Apparatus for estimating sleeping state - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an apparatus for simply estimating a sleeping state in real time with a small amount of calculation. <P>SOLUTION: The apparatus for estimating a sleeping state is provided with: a period detection part for detecting a pulsation period from a metered heart rate; a short period standard deviation calculation part for calculating short period standard deviation from the pulsation period; a long period standard deviation calculation part for calculating long period standard deviation from the pulsation period; an index calculation part for calculating a sleeping state index by putting the short period standard deviation and the long period standard deviation into a prescribed calculation formula; and a sleeping state judgment part for judging the sleeping state of the user based on the calculated sleeping state index. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a sleep state estimation device for estimating a sleep state of a living body from heartbeat data of the living body.

  In recent years, interest in sleep is increasing as health awareness increases. In addition, there is a need to easily measure the sleep state at home, such as the effect of sleep apnea syndrome on daily life becomes a problem.

  It is well known that the sleep state changes periodically, and as a technique for detecting such a change, a polysomnogram that measures brain waves, eye movements, myoelectricity, and the like is known. However, this has a problem that it is difficult to use for estimating a normal sleep state at home because the device is large-scale or many sensors need to be worn on the body.

As an alternative to the polysomnogram, attempts have been made to measure heart rate, respiration, body movement, and the like, and estimate the sleep state using the measurement value of the period and the fluctuation of the period. For example, changes in pulse rate and changes over time are used as indices (Patent Document 1 and Patent Document 2), standard deviation of R-wave interval of heartbeat and parameters obtained by frequency analysis of R-wave interval are used as indexes. And the like (Patent Document 3), and those using eight parameters derived from the heart rate as indices (Patent Document 4).
JP-A-63-283623 JP 63-205592 A JP-A-7-143972 JP 2003-260040 A

  However, the above method requires many complicated calculations in order to obtain an index for sleep state estimation. In Patent Document 3, frequency analysis of heartbeat R waves is performed in order to calculate an index used for estimation. In Patent Document 4, eight kinds of indices are calculated from three signal values from heartbeat, respiration, and body movement. ing. When performing such calculations, it is necessary to use measurements over the past few minutes. For example, in the frequency analysis method, data of at least about 256 to 512 points is used. Taking a heartbeat as an example, if one beat of the heartbeat is 1 second, it is necessary to use a measurement result over 4 to 8 minutes, and real-time characteristics are poor. Further, frequency analysis requires computer power. Thus, the above method has a problem that it is difficult to estimate a sleep state in real time.

  The present invention has been made based on the above circumstances, and an object of the present invention is to provide an apparatus that can easily and in real time estimate a sleep state with a small amount of calculation.

  According to one aspect of the present invention (claim 1), the sleep state estimation apparatus of the present invention detects a pulsation period from a measured heartbeat, and a short-term standard that calculates a short-term standard deviation from the pulsation period. A deviation calculating unit, a long-term standard deviation calculating unit for calculating a long-term standard deviation from the pulsation period, an index calculating unit for calculating a sleep state index by giving the short-term standard deviation and the long-term standard deviation to a predetermined calculation formula; A sleep state determination unit that determines the sleep state of the user based on the calculated sleep state index.

  According to this, since the sleep state index is obtained using two types of standard deviations having different time widths and the sleep state is determined based on the sleep state index, the sleep state can be estimated easily and with high accuracy.

  In the sleep state estimation apparatus according to another aspect of the present invention (claim 2), the predetermined calculation formula includes a function having a sigmoid function characteristic.

  According to this, since the calculation formula includes a function having a sigmoid function characteristic, whether the subject is in a deep sleep state or a shallow sleep state based on two standard deviation values of a short-term standard deviation value and a long-term standard deviation value. The sleep state can be accurately determined by optimizing the function constants.

  In the sleep state estimation apparatus according to another aspect of the present invention (Claim 3), the predetermined calculation formula includes a function having a sigmoid function characteristic into which the short-term standard deviation is substituted, and the long-term standard deviation is substituted. And a sum with a function having a sigmoid function characteristic.

In the sleep state estimation apparatus according to another aspect of the present invention (Claim 4), the sleep state index is SI, A1, A2, B1, B2, M1, M2, k1, k2 are constants, and nLSD is a long term. When the standard deviation and nSSD are short-term standard deviations, the predetermined calculation formula for obtaining the amount of change ΔSI of the sleep state index is:

It is represented by

  In the sleep state estimation apparatus according to another aspect of the present invention (Claim 5), the nLSD is a value obtained by dividing the long-term standard deviation by the average value of the long-term standard deviation, and the nSSD is the short-term standard deviation. It is the value divided by the average value of the short-term standard deviation. According to this, it becomes possible to obtain the sleep state index based on values obtained by normalizing the long-term standard deviation and the short-term standard deviation by their average values.

  In the sleep state estimation apparatus according to another aspect of the present invention (Claim 6), the short-term standard deviation is calculated based on a predetermined beat cycle of 2 or 3 beats, and the long-term standard deviation is 4 to 4. It is calculated based on a predetermined beat cycle of 30 beats.

  In the sleep state estimating apparatus according to another aspect of the present invention (claim 7), the long-term standard deviation is calculated using a moving average value of a pulsation period. According to this, since the moving average value of the pulsation period is used, the influence of the short-term standard deviation can be removed in calculating the long-term standard deviation.

  In the sleep state estimation apparatus according to another aspect of the present invention (Claim 8), the short-term standard deviation is calculated based on a predetermined time period of 2 seconds or 3 seconds, and the long-term standard deviation is 4 to 30. Calculated based on a predetermined time period of seconds.

  In the sleep state estimating apparatus according to another aspect of the present invention (claim 9), the long-term standard deviation is calculated using a moving average value of a time period. According to this, since the moving average value of the time period is used, the influence of the short-term standard deviation can be removed in calculating the long-term standard deviation.

  In the sleep state estimation apparatus according to another aspect of the present invention (claim 10), the sleep state is determined based on a comparison between the calculated sleep state index and a predetermined threshold value.

  The relationship between the heart rate and the sleep state used in the sleep state determination method according to the embodiment of the present invention will be described.

  It is known that the RR interval, which is the cycle of each heartbeat during sleep, is not constant, and varies around an average interval of 1 second as shown in FIG. It is also known that the power spectrum of heart rate fluctuation shows different aspects depending on the state of the autonomic nervous system. That is, when the low frequency component (LF component (0.04Hz-0.15Hz)) is large in the heartbeat, this indicates that the sympathetic nerve is active, and when the high frequency component (HF component (0.15Hz-0.40Hz)) is large, the parasympathy. Indicates that the nerve is active. It is known that when the parasympathetic nerve is active, it is deep sleep, and when the sympathetic nerve is active, it is shallow sleep. Thus, it is known that there is a correlation between the sleep depth and the frequency component of the heart rate.

  In the present invention, the high correlation between the short-term standard deviation and the HF component and the high correlation between the long-term standard deviation and the LF component are used. FIGS. 3A to 3D are graphs for explaining these correlations.

  FIG. 3A is a graph showing a value of a short-term standard deviation described later in time series. The unit of the horizontal axis is “time”, which indicates the passage of time from the start of sleep. Here, the standard deviation for two beats of the RR interval is calculated and graphed. FIG. 3B shows a time-series graph obtained by extracting the HF component (power value at 0.15 Hz-0.40 Hz) of the heartbeat obtained by frequency analysis for 256 beats of the RR interval. Here, the integrated value of the frequency width of 0.15 Hz-0.40 Hz at that time is graphed. FIG. 3C is a graph showing long-term standard deviation values described later in time series. The unit of the horizontal axis is “time”, which indicates the passage of time from the start of sleep. Here, the standard deviation over 10 beats is calculated and graphed for the RR interval taking the 5-beat moving average. FIG. 3D shows an LF component (power value at 0.04 Hz-0.15 Hz) of the heartbeat obtained by frequency analysis for 256 beats of the RR interval, and graphed in time series. Here, the integrated value of the frequency width of 0.04 Hz-0.15 Hz at that time is graphed.

  Comparing the short-term standard deviation (FIG. 3A) and the HF component (FIG. 3B), it can be seen that the graph shapes correspond to each other. The correlation coefficient between them was calculated to be 0.787. In addition, it can be seen that the graph shapes correspond to each other in the comparison of the long-term standard deviation (FIG. 3C) and the LF component (FIG. 3D). The correlation coefficient between the two was 0.903. Thus, a high correlation is recognized between the standard deviation and the frequency component of the heartbeat.

  Even if the number of beats used for frequency analysis is changed, the change in the correlation is not large, but the correlation tends to increase as the number increases. Further, even if the number of beats for calculating the long-term standard deviation is changed, the correlation with LF is kept at a high value, but it is desirable to have up to about 30 beats.

  In general, when the HF component of the RR interval is large, parasympathetic nerve activity is active and sleep is deep. On the other hand, when the LF component of the RR interval is large, it is said that the activity of the sympathetic nerve is active and the sleep is shallow.

  In consideration of the relationship between the sympathetic nervous system and the RR interval of the heartbeat as described above, and the correlation between the frequency analysis result of the RR interval and the standard deviation value, the short-term standard deviation and the long-term standard deviation are used. In the present invention, the sleep state index SI is easily obtained.

  Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is an overall configuration diagram of a sleep state estimation device 100 according to an embodiment of the present invention. FIG. 6 is a flowchart according to an embodiment of the present invention.

  The sleep state estimation device 100 includes a measurement unit 101, a cycle detection unit 102, a short-term standard deviation calculation unit 103, a long-term standard deviation calculation unit 104, a change amount calculation unit 105, a sleep state calculation unit 106, and a sleep state determination unit described below. 107 and a display unit 108.

  The period detection unit 102, the short-term standard deviation calculation unit 103, the long-term standard deviation calculation unit 104, the change amount calculation unit 105, the sleep state calculation unit 106, and the sleep state determination unit 107 perform predetermined calculation processing and determination processing. These can be realized by individual processing apparatuses, but are realized by one computer 110 in the present embodiment. The computer 110 used in the present embodiment includes, for example, processing means such as a CPU (not shown), storage means such as ROM, RAM, and hard disk drive, and input means such as a keyboard and a mouse. Further, the storage means stores a program for realizing the above-described units. The main parts of the sleep state estimation apparatus 100 of the present invention are the short-term standard deviation calculation unit 103 to the sleep state determination unit 107, and the measurement unit 101, the cycle detection unit 102, and the display unit 108 are essential constituent requirements of the present invention. is not. For example, the cycle detection unit 102 may be provided outside the sleep state estimation device 100 and supply data detected by the cycle detection unit 102 by a known method to the short-term standard deviation calculation unit 103.

  The measurement unit 101 is connected to the computer 110 via, for example, a USB, a serial bus, or a parallel bus. The display unit 108 is connected to the computer 110 via, for example, a D-sub connector.

  The measuring unit 101 has a function of detecting a heartbeat or a pulse and converting it into an electrical signal to acquire heartbeat data. Specifically, the electrocardiograph is attached to a human body and acquires a heart rate. In the present embodiment, the sampling rate is 200 (Hz). The measuring unit 101 measures heartbeats from the subject via the electrocardiograph, sends the measured heartbeats to the computer 101, and sequentially stores them in the storage means (S601). The acquired heartbeat data is sent to the cycle detection unit 102 and sequentially stored in the storage means of the cycle detection unit 102. As an alternative method of acquiring heart rate data, a pulse may be acquired from a pulse wave sensor worn by a human. This is a pulse wave sensor attached to a subject's hand with a band, and a light emitting element and a light receiving element are arranged on the fingernail side and the back side, respectively, and a blood flow is detected to obtain a signal corresponding to the pulse. . In the present embodiment, heart rate data obtained from an electrocardiograph will be described as an example.

  The cycle detection unit 102 detects the cycle (RR interval) for each beat from the heartbeat data stored in the storage unit, and stores it in the storage unit (S602). The period for each beat is, for example, the time interval between the vertices of the obtained heartbeat data.

  The short-term standard deviation calculator 103 calculates a short-term standard deviation (hereinafter referred to as SSD) of the RR interval from the heart rate RR interval data stored in the storage means (S603). In general, the range of the short-term frequency (HF) of the RR interval is 0.15 to 0.4 Hz, which is considered to be 1.25 to 3 in consideration of the fact that 1/2 period is considered in Fourier analysis. The period is 3 seconds. Therefore, in order to extract the fluctuation of the heart rate limited to the range of HF, it is reasonable to calculate the SSD of 1.25 to 3.3 seconds (2 or 3 beats). Considering that the heart rate is generally about 1 beat / second during sleep, in this embodiment, the “short-term” section is set to 2 beats (about 2 seconds). The SSD was calculated using the RR interval.

  Further, the short-term standard deviation calculating unit 103 obtains an average value of the SSD from the start of measurement, and calculates nSSD that is a value obtained by dividing the SSD by this average value. Then, the calculated nSSD is sent to the change amount calculation unit 105. The nSSD, which is the value obtained by dividing the SSD by the average value of the SSD from the start of measurement, is used to calculate the sleep state index. The accuracy of sleep state estimation is reduced by reducing individual differences in heart rate variability. It is for improving. However, the calculation amount may be reduced by directly calculating the SSD without calculating the nSSD.

  On the other hand, the long-term standard deviation calculation unit 104 obtains the moving average value of the RR interval from the RR interval data of the heartbeat stored in the storage means, and the long-term standard deviation from the calculated moving average value of the RR interval. (Hereinafter referred to as LSD) is calculated (S604).

  Here, the moving average value is an average value obtained continuously while moving the average value of a certain number of data on the time axis. The reason why the moving average value is calculated is to remove the influence of the short-term standard deviation when calculating the LSD. In this embodiment, a moving average value of 5 beats (about 5 seconds) is obtained. The interval of 5 beats is determined in correspondence with the short-term frequency (HF) of the RR interval. That is, if the LSD is simply calculated, a short-term variation is mixed in the calculation result. Therefore, in this embodiment, a method is adopted in which short-term variation is first reduced by a moving average for five beats, and then LSD is calculated. The number of beats (5 beats in this embodiment) at the time of calculating the moving average is a value that can sufficiently eliminate the influence of SSD in consideration of short-term and long-term values (2 beats and 10 beats in this embodiment, respectively). Should be set. However, it is essential to calculate the LSD at intervals longer than the number of beats used for the moving average.

  Next, the LSD section will be described. In general, the range of the long-term frequency (LF) of the RR interval is 0.04 to 0.15 Hz, which is 3.3 to 12 in consideration of the fact that 1/2 period is considered in Fourier analysis. The cycle is 5 seconds (about 4 to 12 beats). However, when the standard deviation of LF and long-term (4 to 500 beats) was actually compared, it was found that the correlation coefficient up to about 30 beats exceeded 0.8. Therefore, in the present invention, when extracting the fluctuation of the heart rate limited to the range of LF, it has been determined that it is reasonable to calculate the LSD of the section of 4 to 30 beats. However, it is also possible to estimate the sleep state by calculating LSD with the number of beats of 31 beats or more. It is desirable to determine the number of beats in consideration of the accuracy required for the estimation result and the required real-time property.

  In this embodiment, in order to clarify the division from the short-term standard deviation (SSD), after calculating the moving average every 5 beats, the “long-term” division is set to 10 beats (about 10 seconds). The LSD was calculated using the RR interval for 10 beats. The heart rate during sleep is generally about 1 beat / second.

  The long-term standard deviation calculation unit 104 has a function of obtaining an average value of LSD from the start of measurement and calculating nLSD that is a value obtained by dividing the LSD by this average value. Then, the calculated nLSD is sent to the change amount calculation unit 105. Here again, nLSD, which is a value obtained by dividing LSD by the average value of the long-term standard deviation from the start of measurement, is used to calculate the sleep state index. This is to improve the accuracy of state estimation. However, the calculation amount may be reduced by directly calculating LSD without calculating nLSD.

  The calculation of the short-term standard deviation and the long-term standard deviation may be performed in parallel to obtain both standard deviations simultaneously.

The change amount calculation unit 105 calculates the change value ΔSI (t) per unit time of the sleep state index at time t by substituting the value of nSSD and the value of nLSD into the following calculation formula (Formula 1). (S605).

  In (Formula 1), SI (t-1) is the immediately preceding sleep state index. A1, A2, B1, B2, M1, M2, k1, and k2 are constants that are determined depending on the calculation interval of the sleep state index SI. In this example, the sleep state index SI is calculated at a time interval of 5 seconds, and A1 = 0.7, A2 = 0.07, B1 = 95, B2 = 40, M1 = M2 = 1.3, Use k1 = 10 and k2 = 4.

  Here, the calculation interval of the sleep state index SI is not limited to 5 seconds, and may be appropriately determined according to the computing capability of the computer 110 and the output cycle of the required estimation result. When increasing the calculation interval of the sleep state index SI, the interval may be about 30 seconds. This is because a change from deep sleep to shallow sleep may occur in about 30 seconds when the change in BIS value is observed. However, even if the estimated value is calculated with a time width exceeding this, the possibility of inaccuracy increases. Since the constants A1 and A2 in the above (Equation 1) are parameters for obtaining the amount of change per unit time, when the calculation time width of the sleep state index SI is changed, the values of A1 and A2 also need to be changed.

  The first term of (Equation 1) is a factor that pushes up the sleep state index SI by a long-term standard deviation value (LSD) corresponding to the LF component that is an index of sympathetic nerve activity, and the second term of (Equation 1) is The short-term standard deviation value (SSD) corresponding to the HF component serving as an index of parasympathetic nerve activity is used as a factor for depressing the sleep state index SI. As another embodiment, nSSD / nLSD may be used instead of nSSD.

  In addition, the above (Equation 1) is easy to understand which value (M1, M2) is the center and what range of numerical values (k1, k2) is reflected in the change ΔSI (t), and constants A1 to k2 There is an advantage that it is easy to adjust to fit the actual sleep state. Here, the constants A1 and A2 determine the amount of change per unit estimated time, and the constants B1 and B2 determine the upper and lower limits of the sleep state index SI. The constants A <b> 1 to k <b> 2 are not limited to those described above, and other combinations of constants that are appropriately reset according to the characteristics of the subject whose sleep state is estimated may be used.

  The sleep state index SI in the present embodiment is based on a well-known BIS (Bispectral index) monitor that measures a sleep state based on an electroencephalogram or the like, so that the sleep state can be expressed on a scale from 100 (wakefulness) to 0. It has become.

  (Equation 1) is set so that a significant change in ΔSI occurs only for nLSD and nSSD values within a predetermined range, and shows a substantially constant value before and after that, and has a sigmoid function characteristic. Is. However, the calculation formula of the sleep state index SI used in one embodiment of the present invention is not limited to the function as in the above formula, and a mathematical formula or algorithm having similar characteristics can also be used.

  The change amount calculation unit 105 sends the calculated change value ΔSI (t) to the sleep state calculation unit 106.

  The sleep state calculation unit 106 calculates the sleep state index SI (t) at time t (S606). More specifically, the current sleep state index SI (t) is calculated by adding the sent change value ΔSI (t) to the immediately preceding sleep state index SI (t−1). Further, the calculated sleep state index SI (t) is converted into display data such as a graph so as to be referred to by the examiner, and is sent to the display unit 108. Further, in order to determine the sleep state, the calculated sleep state index SI is also sent to the sleep state determination unit 107. Further, the calculated sleep state index SI is sequentially stored in the storage means so that the past sleep state index can be referred to later.

  The sleep state determination unit 107 compares the sleep state index SI with a predetermined threshold to determine the current sleep state (whether it is deep sleep or light sleep) (S607). In the present embodiment, deep sleep refers to sleep stage 3 and sleep stage 4 out of six stages of sleep, which are international standards, and shallow sleep refers to sleep stage 1 and sleep stage 2. The relationship between sleep stages and BIS values according to the well-known polysomnographic method is as follows: wakefulness state: 93.9 ± 2.1, sleep stage 1: 88.3 ± 3.9, sleep stage 2: 80.5 ± 4.4, sleep stage 3: 54.9 ± 6.1, sleep stage 4 : 40.9 ± 6.3 (Takahashi et al., Anesth. Analg., 1998 vol.86, S234).

  In the present embodiment, the predetermined threshold is set to 70, and when the sleep state index SI is equal to or greater than the threshold, it is determined that the sleep is shallow, and when it is small, it is determined that the sleep is deep. The determination result is converted into predetermined display data and sent to the display unit 108. In addition, it is good also as a structure which determines the sleep state of more steps using a some threshold value.

  The display unit 108 is, for example, a display, and is connected to a computer 110 including a sleep state calculation unit 106 and a sleep state determination unit 107. The display unit 108 displays the sleep state index subjected to predetermined image processing such as conversion to display data and the sleep state determination result (S608). As the display unit 108, a printing device such as a USB or a network-connected printer can be added.

  The result of the sleep state index SI calculated by (Equation 1) which is one embodiment of the present invention is shown in FIG. Moreover, the sleep state index measured by conventional BIS is shown in FIG. The correlation coefficient between the sleep state index measured by the conventional method BIS and the sleep state index calculated from the two standard deviations of the heart rate (FIG. 5) was 0.677.

  The sleep state determination unit 107 determines the sleep depth from the obtained sleep state index, and the determination is based on the sleep stage criteria of Takahashi et al. (Anesth, Analg., 1998 vol.86, S234) described above. ing. In this embodiment, the sleep state index 70 or more is shallow sleep (sleep stages 1 and 2), and less than 70 is deep sleep (sleep stages 3 and 4). As a result, the correct answer rate for sleep state determination was 82.2%.

  As described above, the sleep state index is obtained using two types of standard deviations having different time widths, and the sleep state is determined based on the sleep state index. Therefore, according to the present invention, the sleep state can be easily and accurately determined. Can be estimated.

  Moreover, according to this method, since a sleep determination result can be obtained in real time, the subject can be awakened when it is determined that sleep is shallow, and a comfortable awakening can be applied. Moreover, since this method does not perform frequency analysis in the calculation process, the calculation cost can be reduced. Moreover, since it is only necessary to obtain heart rate data from the electrocardiograph, the sleep state can be estimated with a small number of sensors.

  Furthermore, since the two types of standard deviations used in this method are highly correlated with the HF and LF components used as the activity index of the autonomic nervous system by the frequency analysis of heart rate variability, this method is only for estimating the sleep state. It can also be used to estimate the degree of arousal and relaxation.

  In this embodiment, the short-term standard deviation and the long-term standard deviation are calculated based on the number of beats on the assumption that the heart rate during sleep is generally about 1 beat / second. However, in practice, 1 beat = 1 second is not always the case, and the peak of the electrocardiogram measurement rarely appears at a perfect timing. For this reason, there may be a case where the time does not coincide with the calculation process (Equation 1) of the sleep state index SI performed every predetermined number of seconds, and in this case, a device for the calculation process as described below is required. Please note that.

  For example, if the peak appearance time is 22.850 seconds, 23.775 seconds, 24.675 seconds, and 28.585 seconds after the start of measurement, the RR interval at time 23.775 seconds is 0.925 seconds, and the RR interval at time 24.675 seconds is The RR interval at 0.900 seconds and time 25.585 seconds is 0.910 seconds. Then, a standard deviation is calculated from successive RR intervals. That is, the short-term standard deviation at time 24.675 seconds is 0.0177, and the short-term standard deviation at time 25.585 seconds is 0.0071. The moving average for 5 beats and the long-term standard deviation for 10 beats are similarly calculated.

  On the other hand, the sleep state index SI is calculated every predetermined time interval (5 seconds in this embodiment). For this reason, it does not agree with the time transition in the calculation of the short-term standard deviation and the long-term standard deviation. In order to solve this problem, in this embodiment, a process of taking a value immediately before the time is performed. With this process, in the above example, the short-term standard deviation at time 25 seconds is defined as 0.0177. Also, a known interpolation process may be performed. If linear interpolation is used in the above example, the short-term standard deviation x = 0.0110 at the time of 25 seconds is obtained from the relationship (25-24.675) :( 25.585-24.675) = (x-0.0071): (0.0177-0.0071) .

In the present embodiment, the short-term standard deviation and the long-term standard deviation are calculated based on the number of beats, but the short-term standard deviation and the long-term standard deviation may be calculated using a predetermined number of seconds instead of the number of beats. In this case, for example, a short-term standard deviation of 2 seconds of the RR interval is calculated, and a long-term standard deviation over 10 seconds is calculated for the RR interval obtained by taking a 5-second moving average.

  As described above, when calculating the short-term standard deviation and the long-term standard deviation based on the number of seconds instead of the number of beats, in order to match the timing of the calculation of the standard deviation and the calculation of the sleep state index SI, (1) RR interval is extracted from the heartbeat peak that appears during the scheduled time, and short-term standard deviation is calculated based on the extracted RR interval. (2) RR interval or standard deviation immediately before the scheduled time, or the above interpolation processing Various methods such as using a value obtained by performing the above are applicable.

The schematic diagram of the sleep state estimation apparatus according to one embodiment of this invention. The figure showing the value of heart rate RR interval during sleep according to one embodiment of this invention. The figure showing the comparison of short-term standard deviation and HF, long-term standard deviation, and LF according to one Embodiment of this invention. The figure showing the sleep state index | exponent obtained by the BIS monitor. The figure showing the sleep state index calculated from the heartbeat according to one embodiment of this invention. The flowchart figure of the sleep state estimation method according to one Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Measurement part 102 Period detection part 103 Short-term standard deviation calculation part 104 Long-term standard deviation calculation part 105 Change amount calculation part 106 Sleep state calculation part 107 Sleep state determination part

Claims (12)

A cycle detection unit that detects a pulsation cycle from the measured heartbeat;
A short-term standard deviation calculator for calculating a short-term standard deviation from the pulsation cycle;
A long-term standard deviation calculator for calculating a long-term standard deviation from the pulsation cycle;
An index calculation unit for calculating a sleep state index by giving the short-term standard deviation and the long-term standard deviation to a predetermined calculation formula;
A sleep state determination unit that determines the sleep state of the user based on the calculated sleep state index;
A sleep state estimating device.   The sleep state estimation apparatus according to claim 1, wherein the predetermined calculation formula includes a function having a sigmoid function characteristic.   The sleep according to claim 1, wherein the predetermined calculation formula includes a sum of a function having a sigmoid function characteristic into which the short-term standard deviation is substituted and a function having a sigmoid function characteristic into which the long-term standard deviation is substituted. State estimation device. When the sleep state index is SI, A1, A2, B1, B2, M1, M2, k1, and k2 are constants, nLSD is a long-term standard deviation, and nSSD is a short-term standard deviation, the amount of change ΔSI of the sleep state index is The predetermined calculation formula for obtaining is

The sleep state estimation apparatus according to claim 1, represented by:   The sleep state according to claim 4, wherein the nLSD is a value obtained by dividing the long-term standard deviation by an average value of the long-term standard deviation, and the nSSD is a value obtained by dividing the short-term standard deviation by the average value of the short-term standard deviation. Estimating device.   The short-term standard deviation is calculated based on a predetermined beat cycle of 2 or 3 beats, and the long-term standard deviation is calculated based on a predetermined beat cycle of 4 to 30 beats. Sleep state estimation device.   The sleep state estimation apparatus according to claim 6, wherein the long-term standard deviation is calculated using a moving average value of the pulsation period.   The sleep state according to claim 1, wherein the short-term standard deviation is calculated based on a predetermined time period of 2 seconds or 3 seconds, and the long-term standard deviation is calculated based on a predetermined time period of 4 to 30 seconds. Estimating device.   The sleep state estimation apparatus according to claim 8, wherein the long-term standard deviation is calculated using a moving average value of the time period.   The sleep state estimation apparatus according to claim 1, wherein the determination of the sleep state is performed based on a comparison between the calculated sleep state index and a predetermined threshold value. Detecting a beat cycle from the heartbeat;
Calculating a short-term standard deviation from the beat period;
Calculating a long-term standard deviation from the beat period;
Giving a short-term standard deviation and a long-term standard deviation to a predetermined calculation formula to calculate a sleep state index;
Determining the sleep state of the user based on the calculated sleep state index;
A sleep state estimation method including: On the computer,
A cycle detection function that detects the pulse cycle from the heartbeat;
A short-term standard deviation calculation function for calculating a short-term standard deviation from the pulsation cycle;
A long-term standard deviation calculating function for calculating a long-term standard deviation from the pulsation period;
An index calculation function for calculating a sleep state index by giving the short-term standard deviation and the long-term standard deviation to a predetermined calculation formula;
A sleep state determination function for determining the sleep state of the user based on the calculated sleep state index;
A program to realize

Images